Sam 𝕏u

(1/2) Check out 𝐌𝐞𝐬𝐡𝐆𝐏𝐓! MeshGPT generates triangle meshes by autoregressively sampling from a transformer model that produces tokens from a learned geometric vocabulary. As a result, we obtain clean and compact meshes :) nihalsid.github.io/mesh-gpt/ youtu.be/UV90O1_69_o

Two PhD students graduated two weeks ago: Yuzhe Qin @QinYuzhe (co-advised with Hao Su), and Yueh-Hua Wu @yh_kris. They are my first batch of robotics students. When I was a student, Alyosha told me: "Good students are your friends, you can learn from them." Yuzhe and Yueh-Hua helped me build a robotics lab when I only knew about computer vision. I have learned so much along the journey. Yuzhe is becoming a fancy CTO of his own robotics startup, DexMate. He single-handed all the robotics hardware and simulation setup in our lab for a long period of time. You probably don't need many people to do robotics, you only need one Yuzhe. He has been leading the direction of dexterous manipulation from when no one cares to now everyone is doing it. And he is naming his company after it. Yueh-Hua is joining Nvidia as a research scientist. Yueh-Hua also contributed to our initial efforts on dexterous manipulation, focusing on the Reinforcement Learning part. He later developed his own interests in offline RL, and using foundation models for control. He works on connecting 3D visual representation learning to control very ahead of time. I am sure Nvidia will be a good place to do all this research.

Thank you @_akhaliq for sharing our work! In this #CVPR2024 project, we introduce Video2Game to convert a video capture of a real-world scene into an interactive game. video2game.github.io